The group works on a number of research areas including:
- High-performance MPI implementation
- Fault Tolerance and Resilience
- Heterogeneous computing and memory systems
- Lightweight Communication and Threading Models
MPICH is the flagship project in the group, and is one of the most widely used implementations of MPI in the world. It forms the basis of various partner MPI implementations including Intel MPI, IBM MPI, Cray MPI, Microsoft MPI, MVAPICH, and various others. Other projects in the group include the Data Movement for Heterogeneous Memory (DMEM) project that deals with improving the usability of non-DRAM based memory regions including byte-addressable NVRAM, scratchpad memory, compute-capable memory, accelerator memory, and various other types of memory. The group is also involved in the Global View Resilience (GVR) project that deals with memory resilience using multi-versioned memory snapshots and cross-layer error information sharing, and the ARGO project that focuses on designing a new Operating and Low-level Runtime System for Exascale.
Details of the various projects the group is involved in are listed here.
The group is supported by funding from the DOE Advanced Scientific Computing Research (ASCR), DOE National Nuclear Security Administration (NNSA), and National Science Foundation (NSF).